In [1]:
# Import packages
import pandas as pd
import plotly.express as px
import plotly.graph_objs as go
import matplotlib.pyplot as plt
import matplotlib as mpl
import numpy as np
import seaborn as sns

Exploratory analysis¶

In [2]:
Fall2019 = pd.read_csv('RecCen2019.csv')
In [3]:
Fall2022 = pd.read_csv('RecCen_Fall2022 .csv')
In [4]:
Fall2022.head()
Out[4]:
Date Time Visits Day
0 9/18/22 Time 0 Sunday
1 9/18/2022 6:00 a.m. 0 Sunday
2 9/18/2022 7:00 a.m. 0 Sunday
3 9/18/2022 8:00 a.m. 0 Sunday
4 9/18/2022 9:00 a.m. 0 Sunday
In [5]:
# Remove the first row
Fall2022 = Fall2022.tail(-1)
In [6]:
Fall2022.head()
Out[6]:
Date Time Visits Day
1 9/18/2022 6:00 a.m. 0 Sunday
2 9/18/2022 7:00 a.m. 0 Sunday
3 9/18/2022 8:00 a.m. 0 Sunday
4 9/18/2022 9:00 a.m. 0 Sunday
5 9/18/2022 10:00 a.m. 0 Sunday
In [7]:
Fall2019.head()
Out[7]:
Date Time Visits Day
0 9/29/2019 5:00 a.m. 0 Sunday
1 9/29/2019 6:00 a.m. 0 Sunday
2 9/29/2019 7:00 a.m. 0 Sunday
3 9/29/2019 8:00 a.m. 4 Sunday
4 9/29/2019 9:00 a.m. 225 Sunday
In [8]:
Fall2022.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1428 entries, 1 to 1428
Data columns (total 4 columns):
 #   Column  Non-Null Count  Dtype 
---  ------  --------------  ----- 
 0   Date    1428 non-null   object
 1   Time    1428 non-null   object
 2   Visits  1428 non-null   int64 
 3   Day     1428 non-null   object
dtypes: int64(1), object(3)
memory usage: 44.8+ KB
In [9]:
Fall2019.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1386 entries, 0 to 1385
Data columns (total 4 columns):
 #   Column     Non-Null Count  Dtype 
---  ------     --------------  ----- 
 0   Date       1386 non-null   object
 1   Time       1386 non-null   object
 2   Visits     1386 non-null   int64 
 3   Day        1386 non-null   object
dtypes: int64(1), object(3)
memory usage: 43.4+ KB
In [10]:
Fall2022.columns
Out[10]:
Index(['Date', 'Time', 'Visits', 'Day'], dtype='object')
In [11]:
Fall2019.columns
Out[11]:
Index(['Date     ', 'Time     ', 'Visits ', 'Day'], dtype='object')
In [12]:
# Rename columns so there is no weird spaces after
Fall2019 = Fall2019.rename(columns = {'Visits ': 'Visits', 'Date     ': 'Date', 'Time     ': 'Time'})
In [13]:
Fall2022.isnull().sum()
Out[13]:
Date      0
Time      0
Visits    0
Day       0
dtype: int64
In [14]:
Fall2019.isnull().sum()
Out[14]:
Date      0
Time      0
Visits    0
Day       0
dtype: int64

Take out all the values where there is 0 visits, this accounts for how the hours are different for weekdays compared to weekends. For example 6-8 AM on sunday all have 0 visits even though the rec center is not open during that time which would throw off the mean since its counted as an (n) value but it adds 0 visits, also accounts for holidays where rec center is not open¶

In [15]:
Fall2022 = Fall2022[Fall2022.Visits != 0]
Fall2019 = Fall2019[Fall2019.Visits != 0]
In [16]:
Fall2022.groupby(['Day']).sum()
Out[16]:
Visits
Day
Friday 22334
Monday 36363
Saturday 17219
Sunday 20664
Thursday 29676
Tuesday 35830
Wednesday 32497
In [17]:
Fall2019.groupby(['Day']).sum()
Out[17]:
Visits
Day
Friday 24129
Monday 34993
Saturday 16293
Sunday 17745
Thursday 30478
Tuesday 36794
Wednesday 32876

Comparison between Fall 2019 and Fall 2022¶

In [18]:
# Make days in right order
day_order = ['Monday', 'Tuesday', 'Wednesday', 'Thursday', 'Friday', 'Saturday', 'Sunday']

# For 2019 
Fall2019 = Fall2019.reset_index()
Fall2019 = Fall2019.set_index('Day').loc[day_order]
Fall2019 = Fall2019.reset_index()
# Same for 2022
Fall2022 = Fall2022.reset_index()
Fall2022 = Fall2022.set_index('Day').loc[day_order]
Fall2022 = Fall2022.reset_index()
# Group by day and get sums 
days2019 = Fall2019.groupby(['Day']).sum()
days2022 = Fall2022.groupby(['Day']).sum()

# Make plots
fig = go.Line(x = Fall2022.Day.unique() , y = days2022['Visits'], name = 'Fall 2022')
fig2 = go.Line(x = Fall2019.Day.unique(), y = days2019['Visits'], name = 'Fall 2019')

# Overlay them
data = [fig2,fig]
layout = go.Layout(paper_bgcolor='rgba(0,0,0,0)', plot_bgcolor='rgba(0,0,0,0)', barmode = 'overlay')
fig3 = go.Figure(data = data, layout = layout)

# Add titles
fig3.update_layout(
    title="Comparison between Fall 2019 and Fall 2022 by day",
    xaxis_title="Day",
    yaxis_title="Total amount of visits",
    legend_title="Year",
    font=dict(
        size=14,
        color="Grey"
    )
)

# Show
fig3.show()

/Users/jakejensema/opt/anaconda3/lib/python3.9/site-packages/plotly/graph_objs/_deprecations.py:378: DeprecationWarning: plotly.graph_objs.Line is deprecated.
Please replace it with one of the following more specific types
  - plotly.graph_objs.scatter.Line
  - plotly.graph_objs.layout.shape.Line
  - etc.

  warnings.warn(

We can see the trend is identical for both 2019 and 2022, what about for by time instead of day?¶

In [19]:
# Sort so times are in order 
time_order = ['6:00 a.m.', '7:00 a.m.', '8:00 a.m.', '9:00 a.m.', '10:00 a.m.', '11:00 a.m.', '12:00 p.m.',
             '1:00 p.m.', '2:00 p.m.', '3:00 p.m.', '4:00 p.m.', '5:00 p.m.', '6:00 p.m.', '7:00 p.m.', '8:00 p.m.',
             '9:00 p.m.', '10:00 p.m.']

# For 2019 
Fall2019 = Fall2019.reset_index()
Fall2019 = Fall2019.set_index('Time').loc[time_order]
Fall2019 = Fall2019.reset_index()
# Same for 2022
Fall2022 = Fall2022.reset_index()
Fall2022 = Fall2022.set_index('Time').loc[time_order]
Fall2022 = Fall2022.reset_index()

# Make overlaying plot

# Group by time to and get sum
times2019 = Fall2019.groupby(['Time']).sum()
times2022 = Fall2022.groupby(['Time']).sum()

# Make 2 plots one for 2019 and one for 2022
fig_1 = go.Line(x = Fall2022.Time.unique() , y = times2022['Visits'], name = 'Fall 2022')
fig_2 = go.Line(x = Fall2019.Time.unique(), y = times2019['Visits'], name = 'Fall 2019')

# Overlay them
data = [fig_2,fig_1]
layout = go.Layout(paper_bgcolor='rgba(0,0,0,0)', plot_bgcolor='rgba(0,0,0,0)', barmode = 'overlay')
fig = go.Figure(data = data, layout = layout)

# Add titles
fig.update_layout(
    title="Comparison between Fall 2019 and Fall 2022 by day",
    xaxis_title="Time",
    yaxis_title="Total amount of visits",
    legend_title="Year",
    font=dict(
        size=14,
        color="Grey"
    )
)

# Show
fig.show()

/Users/jakejensema/opt/anaconda3/lib/python3.9/site-packages/plotly/graph_objs/_deprecations.py:378: DeprecationWarning:

plotly.graph_objs.Line is deprecated.
Please replace it with one of the following more specific types
  - plotly.graph_objs.scatter.Line
  - plotly.graph_objs.layout.shape.Line
  - etc.

We can see a very similar pattern for Fall 2019 and Fall 2022 so it would not hurt to combind the data so we have more data to use. Which will help improving the accuracy of our data¶

In [20]:
frames = [Fall2019, Fall2022]

df_new = pd.concat(frames)
In [21]:
plt.style.use('seaborn-dark-palette')
# Take out 5 AM from the dataset because the rec center is not open at 5 AM
df_new = df_new.loc[df_new['Time'] != '5:00 a.m.']

# Get total amount of visits by day
df_day = pd.DataFrame(df_new.groupby('Day').sum())
df_day = df_day.reset_index()

# Set days in order for index
day_order = ["Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday", "Sunday"]

# Set inde=x
df_day = df_day.set_index('Day').loc[day_order]
df_day = df_day.reset_index()

# Make fig
fig = plt.figure(figsize=(9,7))
plt.bar(df_day['Day'], df_day['Visits'])
plt.title('Total amount of people that came into Rec Center by day')

# Show 
plt.show()
In [22]:
# Get data sorted
plt.style.use('seaborn-dark-palette')

# Take out 5 AM from the dataset because the rec center is not open at 5 AM

Visits = df_new.groupby('Time').mean()

# Get total amount of visits by time
df_time = pd.DataFrame(df_new.groupby('Time').sum())
df_time = df_time.reset_index()

# Set new index
df_time = df_time.set_index('Time').loc[time_order]
df_time = df_time.reset_index()

# Make fig
plt.figure(figsize=(9,7))
plt.bar(df_time['Time'], df_time['Visits'])
plt.xlabel('Time', fontsize = 15)
plt.ylabel('Total amount of visits', fontsize = 13)
plt.xticks(fontsize = 11, rotation = 50)
plt.title('Total amount of people that came into Rec Center by hour')

# Show
plt.show()

Charts for article¶

Chart displaying the best time to go to the rec cen¶

In [23]:
# Gradient Color Bar Plots
from matplotlib import cm
from matplotlib.colors import ListedColormap, LinearSegmentedColormap
from matplotlib import colors as mcolors, path
In [24]:
# Get data in correct form to be plotted
df_time_mean = pd.DataFrame(df_new.groupby('Time').mean())
df_time_mean = df_time_mean.reset_index()

df_time_mean = df_time_mean.set_index('Time').loc[time_order]
df_time_mean = df_time_mean.reset_index()

# Plot 
fig, ax = plt.subplots()
bars = ax.bar(df_time_mean['Time'], df_time_mean['Visits'], color = 'blue')

# Make plot Title/Tick marks
plt.title('Average amount of people coming into rec cen per hour', fontsize = 13, weight = 'bold')
plt.xticks(fontsize = 10, rotation = 70)
plt.yticks(fontsize = 10,)

# Change style of graph
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.spines['left'].set_visible(False)
ax.spines['bottom'].set_color('#DDDDDD')
ax.tick_params(bottom=False, left=False)
ax.set_axisbelow(True)
ax.yaxis.grid(True, color='w', linestyle = '--')
ax.xaxis.grid(False)

# Add labels to bars for exact number
bar_color = bars[0].get_facecolor()

for bar in bars:
  ax.text(
      bar.get_x() + bar.get_width() / 2,
      bar.get_height() + 3,
      int(round(bar.get_height(),1)),
      horizontalalignment='center',
      color='black',
      weight='heavy',
      fontsize = 8.3
  )

# Apply gradient for bars
def gradientbars(bars,ydata,cmap):
    ax = bars[0].axes
    lim = ax.get_xlim()+ax.get_ylim()
    ax.axis(lim)
    for bar in bars:
        bar.set_facecolor("none")
        x,y = bar.get_xy()
        w, h = bar.get_width(), bar.get_height()
        grad = np.atleast_2d(np.linspace(0,1*h/max(ydata),256)).T
        ax.imshow(grad, extent=[x,x+w,y,y+h], origin='lower', aspect="auto", 
                  norm=cm.colors.NoNorm(vmin=0,vmax=1), cmap=plt.get_cmap(cmap))
        
# Pass chart through function   
gradientbars(bars, df_time_mean['Visits'], 'cool')

# Show
plt.tight_layout()

Chart showing most popular days¶

In [25]:
# Get data in correct form to be plotted
Visits = df_new.groupby('Day').mean()

df_day_mean = pd.DataFrame(df_new.groupby('Day').mean())
df_day_mean = df_day_mean.reset_index()

day_order = ['Monday', 'Tuesday', 'Wednesday', 'Thursday', 'Friday', 'Saturday', 'Sunday']

df_day_mean = df_day_mean.set_index('Day').loc[day_order]
df_day_mean = df_day_mean.reset_index()

# Plot 
fig, ax = plt.subplots()
bars = ax.bar(df_day_mean['Day'], df_day_mean['Visits'], color = 'blue')

# Make plot Title/Tick marks
plt.title('Average amount of people coming into rec cen per day', fontsize = 13, weight = 'bold')
plt.xticks(fontsize = 9, weight = 'bold')
plt.yticks(fontsize = 10,)

# Change style of graph
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.spines['left'].set_visible(False)
ax.spines['bottom'].set_color('#DDDDDD')
ax.tick_params(bottom=False, left=False)
ax.set_axisbelow(True)
ax.xaxis.grid(False)

# Add labels to bars for exact number
bar_color = bars[0].get_facecolor()

for bar in bars:
  ax.text(
      bar.get_x() + bar.get_width() / 2,
      bar.get_height() + 3,
      int(round(bar.get_height(),1)),
      horizontalalignment='center',
      color='black',
      weight='heavy',
      fontsize = 8.3
  )

def gradientbars(bars,ydata,cmap):
    ax = bars[0].axes
    lim = ax.get_xlim()+ax.get_ylim()
    ax.axis(lim)
    for bar in bars:
        bar.set_facecolor("none")
        x,y = bar.get_xy()
        w, h = bar.get_width(), bar.get_height()
        grad = np.atleast_2d(np.linspace(0,1*h/max(ydata),256)).T
        ax.imshow(grad, extent=[x,x+w,y,y+h], origin='lower', aspect="auto", 
                  norm=cm.colors.NoNorm(vmin=0,vmax=1), cmap=plt.get_cmap(cmap))
        
# Pass chart through function         
gradientbars(bars, df_day_mean['Visits'], 'RdYlBu_r')

# Show
plt.tight_layout()

Heatmaps between day, time, and amount of visits.¶

In [26]:
df = df_new.drop(['Date'], axis = 1)

# Set times in order
df['Time'] = df['Time'].replace({'6:00 a.m.': 6, '7:00 a.m.': 7, '8:00 a.m.': 8, '9:00 a.m.': 9,
             '10:00 a.m.': 10, '11:00 a.m.': 11, '12:00 p.m.': 12, '1:00 p.m.': 13, '2:00 p.m.': 14,
             '3:00 p.m.': 15, '4:00 p.m.': 16, '5:00 p.m.': 17, '6:00 p.m.': 18, '7:00 p.m.': 19,
             '8:00 p.m.':20, '9:00 p.m.': 21, '10:00 p.m.': 22})

# Make new dataframe for heatmap
heatmap1_data = pd.pivot_table(df, values='Visits', 
                     index=['Day'], 
                     columns='Time')

heatmap1_data = heatmap1_data.reset_index()

heatmap_data = heatmap1_data.set_index(['Day']).loc[day_order]

# Make plot
fig = sns.heatmap(heatmap_data, cmap="Reds")
fig.tick_params(left=False, bottom=False)

# Show
fig
Out[26]:
<AxesSubplot:xlabel='Time', ylabel='Day'>

Chart is hard to read the data easily. Putting the times on the y axis would make it easy to see what time is the best¶

In [27]:
df = df_new.drop(['Date'], axis = 1)

# Set days in order
df['Day'] = df['Day'].replace({'Monday': 1, 'Tuesday': 2, 'Wednesday': 3, 'Thursday': 4, 'Friday': 5, 'Saturday': 6, 'Sunday': 7})

# Make dataset
heatmap2_data = pd.pivot_table(df, values='Visits', 
                     index=['Time'], 
                     columns='Day')

heatmap3_data = heatmap2_data.reset_index()

heatmap4_data = heatmap3_data.set_index(['Time']).loc[time_order]


# Loop through cmaps to see what looks best

colors = plt.colormaps()
#for i in colors:
    #fig2 = sns.heatmap(heatmap4_data, cmap=i, vmin = 50, vmax = 300)
    #print(i)
    #plt.show()

# Make and display chart
fig3 = sns.heatmap(heatmap4_data, cmap='Blues', vmin = 50, vmax = 300)
fig3.tick_params(left=False, bottom=False)
plt.savefig('Heatmap_RecCen')

Compare Finals / Midterm week to normal weeks¶

In [28]:
frames = [Fall2019, Fall2022]
df_new = pd.concat(frames)
df_new = df_new.drop(columns = ['level_0', 'index'])

# Make dataframes for midterm and finals week for 2019 and 2022
df_midterm_2019 = df_new.loc[(df_new['Date'] == '10/27/2019') | (df_new['Date'] == '10/28/2019') | 
                       (df_new['Date'] == '10/29/2019') | (df_new['Date'] == '10/30/2019') |
                      (df_new['Date'] == '10/31/2019') | (df_new['Date'] == '11/1/2019')]

df_finals_2019 = df_new.loc[(df_new['Date'] == '12/7/2019') | (df_new['Date'] == '12/8/2019') | 
                       (df_new['Date'] == '12/9/2019') | (df_new['Date'] == '12/10/2019') |
                      (df_new['Date'] == '12/11/2019') | (df_new['Date'] == '12/12/2019') |
                           (df_new['Date'] == '12/12/2019')]

df_midterm_2022 = df_new.loc[(df_new['Date'] == '10/23/2022') | (df_new['Date'] == '10/24/2022') |
                            (df_new['Date'] == '10/25/2022') | (df_new['Date'] == '10/26/2022') |
                            (df_new['Date'] == '10/27/2022') | (df_new['Date'] == '10/28/2022')]

df_finals_2022 = df_new.loc[(df_new['Date'] == '12/3/2022') | (df_new['Date'] == '12/4/2022') |
                           (df_new['Date'] == '12/5/2022') | (df_new['Date'] == '12/6/2022') |
                           (df_new['Date'] == '12/7/2022') | (df_new['Date'] == '12/8/2022') |
                           (df_new['Date'] == '12/9/2022')]

frames = [df_midterm_2019, df_midterm_2022, df_finals_2019, df_finals_2022]

test_df = pd.concat(frames)

# Make dataframe that does not include these days 

df_not_midterm_2019 = df_new.loc[(df_new['Date'] != '10/27/2019') | (df_new['Date'] != '10/28/2019') | 
                       (df_new['Date'] != '10/29/2019') | (df_new['Date'] != '10/30/2019') |
                      (df_new['Date'] != '10/31/2019') | (df_new['Date'] != '11/1/2019')]

df_not_finals_2019 = df_new.loc[(df_new['Date'] != '12/7/2019') | (df_new['Date'] != '12/8/2019') | 
                       (df_new['Date'] != '12/9/2019') | (df_new['Date'] != '12/10/2019') |
                      (df_new['Date'] != '12/11/2019') | (df_new['Date'] != '12/12/2019') |
                           (df_new['Date'] != '12/12/2019')]

df_not_midterm_2022 = df_new.loc[(df_new['Date'] != '10/23/2022') | (df_new['Date'] != '10/24/2022') |
                            (df_new['Date'] != '10/25/2022') | (df_new['Date'] != '10/26/2022') |
                            (df_new['Date'] != '10/27/2022') | (df_new['Date'] != '10/28/2022')]

df_not_finals_2022 = df_new.loc[(df_new['Date'] != '12/3/2022') | (df_new['Date'] != '12/4/2022') |
                           (df_new['Date'] != '12/5/2022') | (df_new['Date'] != '12/6/2022') |
                           (df_new['Date'] != '12/7/2022') | (df_new['Date'] != '12/8/2022') |
                           (df_new['Date'] != '12/9/2022')]

frames2 = [df_not_midterm_2019, df_not_midterm_2022, df_not_finals_2019, df_not_finals_2022]

nontest_df = pd.concat(frames2)
test_df.head()
Out[28]:
Time Day Date Visits
4 6:00 a.m. Monday 10/28/2019 141
14 6:00 a.m. Tuesday 10/29/2019 116
25 6:00 a.m. Wednesday 10/30/2019 143
36 6:00 a.m. Thursday 10/31/2019 67
46 6:00 a.m. Friday 11/1/2019 73

Create visual comparing the two¶

In [29]:
# Get data in correct form to be plotted
df_test_time_mean = pd.DataFrame(test_df.groupby('Time').mean())
df_test_time_mean = df_test_time_mean.reset_index()

df_test_time_mean = df_test_time_mean.set_index('Time').loc[time_order]
df_test_time_mean = df_test_time_mean.reset_index()

# Plot 
fig, ax = plt.subplots()
bars = ax.bar(df_test_time_mean['Time'], df_test_time_mean['Visits'], color = 'blue')

# Make plot Title/Tick marks
plt.title('During midterm and finals week', fontsize = 13, weight = 'bold')
plt.xticks(fontsize = 10, rotation = 70)
plt.yticks(fontsize = 10, color = 'w')

# Change style of graph
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.spines['left'].set_visible(False)
ax.spines['bottom'].set_color('#DDDDDD')
ax.tick_params(bottom=False, left=False)
ax.set_axisbelow(True)
ax.yaxis.grid(False)
ax.xaxis.grid(False)

# Add labels to bars for exact number
bar_color = bars[0].get_facecolor()

for bar in bars:
  ax.text(
      bar.get_x() + bar.get_width() / 2,
      bar.get_height() + 3,
      int(round(bar.get_height(),1)),
      horizontalalignment='center',
      color='black',
      weight='heavy',
      fontsize = 8.3
  )

# Apply gradient for bars
def gradientbars(bars,ydata,cmap):
    ax = bars[0].axes
    lim = ax.get_xlim()+ax.get_ylim()
    ax.axis(lim)
    for bar in bars:
        bar.set_facecolor("none")
        x,y = bar.get_xy()
        w, h = bar.get_width(), bar.get_height()
        grad = np.atleast_2d(np.linspace(0,1*h/max(ydata),256)).T
        ax.imshow(grad, extent=[x,x+w,y,y+h], origin='lower', aspect="auto", 
                  norm=cm.colors.NoNorm(vmin=0,vmax=1), cmap=plt.get_cmap(cmap))
        
# Pass chart through function   
gradientbars(bars, df_test_time_mean['Visits'], 'cool')

plt.tight_layout()
plt.savefig('Finals.png')
In [30]:
# Get data in correct form to be plotted
df_nontest_time_mean = pd.DataFrame(nontest_df.groupby('Time').mean())
df_nontest_time_mean = df_nontest_time_mean.reset_index()

df_nontest_time_mean = df_nontest_time_mean.set_index('Time').loc[time_order]
df_nontest_time_mean = df_nontest_time_mean.reset_index()

# Plot 
fig, ax = plt.subplots()
bars = ax.bar(df_nontest_time_mean['Time'], df_nontest_time_mean['Visits'], color = 'blue')

# Make plot Title/Tick marks
plt.title('Not during midterm and finals week', fontsize = 13, weight = 'bold')
plt.xticks(fontsize = 10, rotation = 70)
plt.yticks(fontsize = 10, color = 'w')

# Change style of graph
ax.spines['top'].set_visible(False)
ax.spines['right'].set_visible(False)
ax.spines['left'].set_visible(False)
ax.spines['bottom'].set_color('#DDDDDD')
ax.tick_params(bottom=False, left=False)
ax.set_axisbelow(True)
ax.yaxis.grid(False)
ax.xaxis.grid(False)

# Add labels to bars for exact number
bar_color = bars[0].get_facecolor()

for bar in bars:
  ax.text(
      bar.get_x() + bar.get_width() / 2,
      bar.get_height() + 3,
      int(round(bar.get_height(),1)),
      horizontalalignment='center',
      color='black',
      weight='heavy',
      fontsize = 8.3
  )

# Apply gradient for bars
def gradientbars(bars,ydata,cmap):
    ax = bars[0].axes
    lim = ax.get_xlim()+ax.get_ylim()
    ax.axis(lim)
    for bar in bars:
        bar.set_facecolor("none")
        x,y = bar.get_xy()
        w, h = bar.get_width(), bar.get_height()
        grad = np.atleast_2d(np.linspace(0,1*h/max(ydata),256)).T
        ax.imshow(grad, extent=[x,x+w,y,y+h], origin='lower', aspect="auto", 
                  norm=cm.colors.NoNorm(vmin=0,vmax=1), cmap=plt.get_cmap(cmap))
        
# Pass chart through function   
gradientbars(bars, df_test_time_mean['Visits'], 'cool')

plt.tight_layout()
plt.savefig('NonFinals.png')

Use linear regression model to see if we can get an accuration prediction on number of visitors¶

Scale the data¶

In [31]:
from sklearn.preprocessing import MinMaxScaler

scaler = MinMaxScaler()
In [32]:
df = df_new.copy()
train_data = df
cols = train_data.columns
In [33]:
train_data = pd.get_dummies(df)

Train test split¶

In [34]:
#import train_test_split
from sklearn.model_selection import train_test_split
In [35]:
X = train_data.drop(columns = 'Visits')
y = train_data['Visits']
In [36]:
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=2)

Linear regression model¶

In [37]:
from sklearn.linear_model import LinearRegression
In [38]:
LR = LinearRegression()
In [39]:
LR.fit(X_train,y_train)
Out[39]:
LinearRegression()
In [40]:
pred = LR.predict(X_test)
# Round up
pred = np.around(pred)

Accuracy score¶

In [41]:
from sklearn.metrics import r2_score
from sklearn.metrics import accuracy_score,confusion_matrix

R2 score¶

In [42]:
r2_score(y_test, pred)
Out[42]:
0.6830996042755412

Accuracy score¶

In [43]:
lr_accuracy=accuracy_score(y_test,pred)
lr_accuracy
Out[43]:
0.008583690987124463
In [44]:
data = {'Visits':y_test, 'Prediction': pred}
comp = pd.DataFrame(data)
comp
Out[44]:
Visits Prediction
855 151 184.0
944 174 137.0
1105 24 55.0
20 114 140.0
587 251 231.0
... ... ...
502 164 207.0
769 344 268.0
778 305 273.0
448 61 37.0
728 165 188.0

466 rows × 2 columns

Conclusion¶

We can see the model did a good job with taking the structure of the data, but it was way too centered around the average amount of people and is not able to tell the "peak" hours from the normal ones. A better approach we could use for this would involve looking at forecasting a time series.¶

Using an ARIMA (AutoRegressive Integrated Moving Average Model) which helps with time forecasting to give a better prediction based on the values from previous times¶

In [45]:
# import libraries
from statsmodels.tsa.arima.model import ARIMA

df_arima = df.drop(columns = ['Day', 'Date'])

df_arima['datetime'] = pd.to_datetime(df_arima['Time'])

df_arima['time'] = pd.to_datetime(df_arima['datetime']).dt.time
df_arima = df_arima.drop(columns = ['datetime', 'Time'])

df_arima = df_arima.set_index('time')
model = ARIMA(df_arima, order=(2,1,2))

results = model.fit()

fit = results.fittedvalues

fit = fit.reset_index()

fit3 = fit
fit2 = fit

fit = fit.groupby('time').mean()

fit = fit.rename(columns = {0: 'Visits'})

/Users/jakejensema/opt/anaconda3/lib/python3.9/site-packages/statsmodels/tsa/base/tsa_model.py:471: ValueWarning:

An unsupported index was provided and will be ignored when e.g. forecasting.

/Users/jakejensema/opt/anaconda3/lib/python3.9/site-packages/statsmodels/tsa/base/tsa_model.py:471: ValueWarning:

An unsupported index was provided and will be ignored when e.g. forecasting.

/Users/jakejensema/opt/anaconda3/lib/python3.9/site-packages/statsmodels/tsa/base/tsa_model.py:471: ValueWarning:

An unsupported index was provided and will be ignored when e.g. forecasting.

Seeing how the ARIMA model did¶

In [46]:
fit2 = fit2.rename(columns = {0: 'Visits'})

df_mean = df.groupby('Time').mean()
df_mean = df_mean.reindex(index = time_order)
fit_plot = fit2.groupby('time').mean()
fit3[0] = fit3[0].round().astype(int)
df['Predicted'] = fit3[0]

# Show plot
df_mean.plot()
Out[46]:
<AxesSubplot:xlabel='Time'>

Metric scores¶

In [47]:
r2_score(df['Visits'], df['Predicted'])
Out[47]:
0.3398511516066497
In [48]:
from sklearn.metrics import mean_squared_error
from sklearn.metrics import mean_absolute_error

mse = mean_squared_error(df['Visits'], df['Predicted'])

print("Mean Squared Error:", mse)

Mean Squared Error: 3715.401287553648

MSE = $\frac{1}{n} \Sigma_{i=1}^n({y}-\hat{y})^2$¶

We see a very high Mean squared error( MSE) because of the outliers in our data.¶